An image, a video or even music, everything can be created with artificial intelligence. Tools using this technology are increasingly within the reach of the average user, and while it has great creative potential, it can also be used for malicious or even criminal intent. This is the case of deepfakes —an acronym for deep (deep) and fake (deception)—, which are essentially multimedia content produced using deep learning algorithms, which manage to impersonate the face or voice of a person with a hyper-realistic result.
Although they are also used as a recreational or humorous resource, deepfakes pose a security threat, since they could circumvent biometric passwords or lead to fraud. In extreme cases, a deepfake could be used to extort or attack the honor and integrity of a person under the threat of disseminating a video, although false, equally compromising. But the most common way that we get it is with the spread of hoaxes and false news based on this technology. So how can we verify that a video is real or fake?
A deepfake can appear so plausible that it is quite easy to believe everything it tells us. Despite its impressive results, with a little attention and common sense, we can avoid falling into its traps.
If a video presents a face, but no sound, there is a first reason to be suspicious. Software to create deepfakes use pre-recorded audio from a mobile phone or computer that can be distorted, with nuances different from those of the original voice. You should also be suspicious if the video is only a few seconds long.
A key issue when identifying hoaxes. Due to the number of hours it takes to create each creative, fake news creators often use patterns on which they later superimpose different faces.
If you notice that there are errors in the synchronization of the audio with the movement of the lips, the eyes or the head, or even perceive inconsistencies in the speech, it is more than likely that you are facing a deepfake.
Synthetic faces are almost indistinguishable from real ones and that can give us a false sense of credibility. However, there are certain natural movements that algorithms have not yet been able to imitate. If the main face presents disproportions in size between the head and the body, or the facial expressions do not correspond to the postures it adopts, you may be seeing a deepfake. Look at other aspects such as eye shadows, facial hair or skin smoothness that will help you differentiate what is real from what is fake.
You can tell if a person is a deepfake if they blink unnaturally or forcefully, or if they do so infrequently.
Another aspect of reality that is difficult to reflect is the color of the lips and the shapes of the tongue and teeth of people when they speak.